What would be the recommended number of spare disks in a simple two node MCC (A400 with SAS SSD's) 3 x DS224C at each site? (144 disks)
It's just beacuse going from one per pool to two, eats up 4 disks 😉 And we also need to add an additional RAID group, which also "hurts"...
I know it's the price to pay for MCC 😉
#Spare disk recommendations in a MCC ?
mellow gazelle
mellow gazelle
Stretched MCC (A400 + ATTO Bridges + DS224C shelfs)
From FUSION 😉
...ahh Legacy "fixed" it... so 4 spares is the anwser but this is also shown even with over 400 disks attached.. is that really true? Or am doing it wrong? I just configure the system, and add shelfs up until 9 shelfs per site... still just 2 spares per site..
maiden nimbus
There's a doc on spares best pracitces IIRC. 1 min
sonic sinew
All your SSDs are the same size right? And this metro has support? Stretched Metro is quite old so I hope these SSDs are not too old too. SSDs fail less often but older SAS SSDs don't seem to be as reliable as newer NVMe ones.
First use RAID4 for root-aggrs.
If you would do 33x SSDs per site --> you need 2x RGs (17x RAID-DP & 16x RAID-DP)
- 2x aggrs with 29x data-disks (per site)
--> 58x data-disks in total
--> 8x parity-disks (per site, for data-aggrs)
If you aim at highest possible supported usable size you could optimize it by using aggrs with different sizes per site (1x aggr with 22x and the other with 44x disks).
This would result in:
- 1x aggr with 20x data-disks (per site)
- 1x aggr with 40x data-disks (per site)
--> 60x data-disks in total
--> only 6x parity-disks (per site, for data-aggrs)
site_A:
- 2x SSDs --> aggrRoot_A with 1x RAID-4 RG (local_A)
- 2x SSDs --> aggrRoot_B with 1x RAID-4 RG (mirror_B)
- 44x SSDs --> aggrSSD_A with 2x RAID-DP RGs (local_A)
- 22x SSDs --> aggrSSD_B with 1x RAID-DP RG (mirror_B)
- 1x SSDs --> Spare (local_A)
- 1x SSDs --> Spare (mirror_B)
site_B:
- 2x SSDs --> aggrRoot_B with 1x RAID-4 RG (local_B)
- 2x SSDs --> aggrRoot_A with 1x RAID-4 RG (mirror_A)
- 22x SSDs --> aggrSSD_B with 1x RAID-DP RG (local_B)
- 44x SSDs --> aggrSSD_A with 2x RAID-DP RGs (mirror_A)
- 1x SSDs --> Spare (local_B)
- 1x SSDs --> Spare (mirror_A)
If you have 4h parts replace (or you have cold spares) it should be ok to only have 1x spare per site.
mellow gazelle
Well we don't have as many disks as I entered on FUSION, just wondered when or if it was time to add additional spares... And yes we have support on this for another 3-4 years as far as I recall... we currently have a balanced setup where the two nodes have the same number of disks assigned (shelf no. 3 isn't fully populated yet)... and we of cause have smaller disks for the root aggregate, and in RAID4... and we have one spare per type of disk in each pool... on each site.. much like you describe. I think 28 disks is the max in each RAID group, but if we can help it, we will of cause try to balance it so that the RG members are equal... But as you might know, sometimes a customer makes strange choises when expanding... mostly because of the costs involved... so you may end up with a not so optimal configuration... especially if they add disks at small numbers each time... then you might end up with one large RAID group at say 22 disks, and then you need to add like 11 additional disks... which isn't optimal at all, but the optimal option to add 22 disks isn't variable for the customer... (some just won't listen) 😉 SSD's may be more reliable yet does NetApp really think one disk per pool is OK if you have over 100 disks in each pool? Seems more like yet another FUSION "issue" to me... 🙂
mellow gazelle
https://docs.netapp.com/us-en/ontap/disks-aggregates/determine-number-disks-part...
we don't even use partitions on a MCC 😉